As in, would hurt the poor the most?
If so: just split the payments (after accounting for the cost to run the system) equally between the city's residents. The poor people will be the first to share cars, so getting an equal share of the profits, while splitting the per-vehicle tolls across the occupants, should be able to counter the regressive effect of a per-vehicle toll. And making it adapt to congestion would prevent the "I got no one to share the ride with this late at night, and I couldn't afford to not split the toll" issue, at least with something like Uber or Waymo to fix the "you didn't drive to work at noon in your own car for late shift" aspect.
And I believe that if ride sharing is effectively forced through capping the traffic to non-congested levels, there would be a practical solution to the issue of organizing ride-sharing for the trip back home late at night.
Like, some sort of dynamic bus network using normal cars (Uber Pool w/ having to switch cars at suitable "hub"s either end of the freeway portion of your trip, or even along the way if any of the resulting legs of your journey would be too-low-throughput to meet realistic latency expectations of yours), at least for the time it takes to manufacture better-suited public transit vehicles.
Software for this can happen in at most a couple months to get something that works, so just make the tolls under this system take half a year to activate after getting enacted for the region.